a 9-dof tractor-semitrailer dynamic handling model for...

LUND UNIVERSITY

PO Box 117221 00 Lund+46 46-222 00 00

A 9-DOF Tractor-Semitrailer Dynamic Handling Model for Advanced Chassis ControlStudies

Gäfvert, Magnus; Lindgärde, Olof

2001

Document Version:Publisher's PDF, also known as Version of record

Link to publication

Citation for published version (APA):Gäfvert, M., & Lindgärde, O. (2001). A 9-DOF Tractor-Semitrailer Dynamic Handling Model for AdvancedChassis Control Studies. (Technical Reports TFRT-7597). Department of Automatic Control, Lund Institute ofTechnology (LTH).

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authorsand/or other copyright owners and it is a condition of accessing publications that users recognise and abide by thelegal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private studyor research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portalTake down policyIf you believe that this document breaches copyright please contact us providing details, and we will removeaccess to the work immediately and investigate your claim.

https://portal.research.lu.se/portal/en/publications/a-9dof-tractorsemitrailer-dynamic-handling-model-for-advanced-chassis-control-studies(05e09d66-6185-445a-a75b-983e13b1edca).html

ISSN 0280–5316ISRN LUTFD2/TFRT--7597--SE

DICOSMOS INTERNAL REPORT

Revision 1.0

A 9-DOF Tractor-SemitrailerDynamic Handling Model

for AdvancedChassis Control Studies

Magnus GäfvertOlof Lindgärde†

†Volvo Technological DevelopmentSweden

Department of Automatic ControlLund Institute of Technology, Sweden

December 2001

Document nameINTERNAL REPORTDate of issueDecember 2001

Department of Automatic ControlLund Institute of TechnologyBox 118SE-221 00 Lund Sweden Document Number

ISRN LUTFD2/TFRT--7597--SESupervisorAuthor(s)

Magnus GäfvertOlof Lindgärde

Sponsoring organization

Title and subtitleA 9-DOF Tractor-Semitrailer Dynamic Handling Model for Advanced Chassis Control Studies.(En dynamisk manövermodell med 9 frihetsgrader av en dragbil med semitrailer för studier av avanceradchassireglering).

AbstractA nonlinear dynamic handling model for a tractor-semitrailer combination vehicle is presented in thisreport. The equations of motion are derived from the fundamental equations of dynamics in Euler'sformulation without approximations. The model is modular in the sense that it is easy to change axleconfiguration, tyre model, suspension model, or to add new features. The primary aim of the modelis simulations of handling scenarios with active yaw control, using unilateral braking and possibly tractorrear wheel steering. Other applications of the model may include real-time hardware-in-the-loopsimulations of tractor-semitrailer handling scenarios. The model is formulated as a state-space modelthat may be implemented in standard simulation environments. A Simulink implementation is presented.Simulation results are compared with experiments to validate the model.

Keywords

Classification system and/or index terms (if any)

Supplementary bibliographical information

ISSN and key title0280-5316

ISBN

LanguageEnglish

Number of pages52

Security classification

Recipient’s notes

The report may be ordered from the Department of Automatic Control or borrowed through:University Library 2, Box 3, SE-221 00 Lund, SwedenFax +46 46 222 44 22 E-mail [email protected]

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . 31.2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . 5

2. Model Overview . . . . . . . . . . . . . . . . . . . . . . . . 53. Vehicle configuration . . . . . . . . . . . . . . . . . . . . . 74. Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

4.1 Motion of a point in vehicle coordinates . . . . . . . . . 84.2 Coordinate transformations . . . . . . . . . . . . . . . . 94.3 Motion of a point on the tractor body . . . . . . . . . . 104.4 Motion of a point on the semitrailer body . . . . . . . . 104.5 Motion of a point on the hitch body . . . . . . . . . . . 10

5. Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105.1 Elimination of internal forces and moments . . . . . . 125.2 State equations . . . . . . . . . . . . . . . . . . . . . . . 145.3 External forces . . . . . . . . . . . . . . . . . . . . . . . 155.4 Axle balance equations . . . . . . . . . . . . . . . . . . 17

6. Tire Models . . . . . . . . . . . . . . . . . . . . . . . . . . . 196.1 Linear Tire Model . . . . . . . . . . . . . . . . . . . . . 206.2 Slip Circle Model . . . . . . . . . . . . . . . . . . . . . . 21

7. Wheel Dynamics . . . . . . . . . . . . . . . . . . . . . . . . 228. Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 22

8.1 Real-Time Performance . . . . . . . . . . . . . . . . . . 239. Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

9.1 Lane-Change Maneuver . . . . . . . . . . . . . . . . . . 259.2 Step-Steering Maneuver . . . . . . . . . . . . . . . . . . 259.3 Random Steering Maneuvers . . . . . . . . . . . . . . . 269.4 Comments on the Results . . . . . . . . . . . . . . . . . 26

10. Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 2611. Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . 2712. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 2713. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 2814. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29A. Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . 31B. Simulink Model . . . . . . . . . . . . . . . . . . . . . . . . . 33C. Moments of inertia . . . . . . . . . . . . . . . . . . . . . . . 35D. Validation Results . . . . . . . . . . . . . . . . . . . . . . . 36

D.1 Lane-change maneuver (i) . . . . . . . . . . . . . . . . 36D.2 Lane-change maneuver (ii) . . . . . . . . . . . . . . . . 37D.3 Step-steering maneuver (i) . . . . . . . . . . . . . . . . 38D.4 Step-steering maneuver (ii) . . . . . . . . . . . . . . . . 39D.5 Random steering maneuvers . . . . . . . . . . . . . . . 40

E. Simulation Results . . . . . . . . . . . . . . . . . . . . . . . 41E.1 Lane-Change Maneuver . . . . . . . . . . . . . . . . . . 41E.2 Lane-Change Maneuver with Jack-knifing . . . . . . . 46

1

1. Introduction

This report is part of the DICOSMOS2 project, under the Swedish NUTEK(VINNOVA) Complex Techical Systems Program. DICOSMOS2 is a jointeffort between the Department of Automatic Control (LTH), Mechatron-ics/Department of Machine Design (KTH), the Department of ComputerEngineering (Chalmers), and Volvo Technological Development (VTD). Theproject is aimed at the study of distributed control of safety-critical motionsystems. Part of the DICOSMOS2 is a study on the design of distributedreal-time control systems on vehicles, initiated by VTD. An active yaw-control system for a tractor-semitrailer commercial vehicle, see Figure 1,was selected as a case study. The present work is part of the results fromthe case study. More results are found in [SCG00, CGS00, GSC00, Gäf01].In this report a nonlinear dynamic handling model for a tractor-semitrailercombination vehicle is derived. The equations of motion are derived fromthe fundamental equations of dynamics in Euler’s formulation without ap-proximations. The primary aim of the model is validation simulations ofhandling scenarios with active yaw control, using unilateral braking andpossibly tractor rear wheel steering. The applications of the model arenot limited to this domain, but may prove useful in other areas of tractor-semitrailer handling simulations. The model is formulated as a state-spacemodel that may be implemented in standard simulation environments. ASimulink implementation is presented. Simulation results are comparedwith experiments to validate the model.

Figure 1 A Volvo FH12 tractor-semitrailer vehicle on the Öresund bridge. (Cour-tesy of Volvo Truck Corporation.)

1.1 Motivation

There are many existing models for simulation of tractor-semitrailer dy-namics. They vary in a wide range of complexity from large multi-bodysystem models with hundreds of degrees of freedom (DOFs), to small 3-DOF bicycle models. It is normally desirable to use the smallest possiblemodels that fit a particular purpose, since they have less parameters, re-

3

quire less computational power, thus simulation time, and are easier tounderstand.

In yaw-control applications the lateral, longitudinal, and yaw motionsof the vehicle are of primary interest. These motions are driven mainlyby the tire-road contact forces. The contact forces depend on lateral andlongitudinal motion of the vehicle, and on the vertical contact forces. Thesimplest possible model to use would then be a 4-DOF model: longitudi-nal, lateral, tractor yaw, and semitrailer yaw motion. It is unfortunatelydifficult to include a good description of tire-road vertical forces in such amodel. These vertical forces vary with load transfer that results from iner-tial forces generated by the acceleration of vehicle masses. For commercialheavy vehicles this load transfer is large because of the high location ofthe center of mass (CM). Apart from using static load balance equations tocompute the vertical forces, it is possible to use body accelerations to modelload transfer. That approach was used in the previous work [GSC00].

These models do still not capture the real effects of load transfer. Onreal vehicles the suspension systems have great influence on the load trans-fer. They introduce a phase lag between the planar motion of the vehicleaxles, or unsprung masses, and the vertical tire-road forces. To capture thisphenomenon correctly it is necessary to use a model divided into sprungand unsprung masses, that are connected by the suspension system. Whenadding the sprung masses additional freedoms are introduced. Some mod-els include only the additional roll freedom, which would be sufficient formodeling lateral motion at constant longitudinal velocity. In this work thechoice became to include all heave, roll, and pitch motions of the sprungmasses, thus introducing five additional DOFs (the heave freedom is com-mon for the tractor and the semitrailer). The reason for this is that inclu-sion of pitch motion results in a more accurate description of longitudinalload transfer. The methods used for deriving the equations of motion donot become more difficult to use with these freedoms. The main effect isthat already large expressions become larger. Since computer algebra toolswas used this did not pose any real problems.

The reasons for deriving a new model instead of using an existing areseveral: Many previous models were derived in the 60s and 70s. These werederived by hand and therefore often subject to approximations that are notnecessary with computer-algebra based methods available today. Further-more, some equations were too large to solve analytically, and were thusleft to numerical solution by iterative methods in the model implementa-tions. This increases the computational complexity of the model, and mayseverely slow down simulations. These early models are also published asvery large equations that are error-prone to implement in a simulationenvironment. Still these models provide good sources of knowledge whenderiving a new model. Recent models are often of large complexity withmany DOFs. Popular modeling environments are multi-body system soft-ware such as Adams. These often result in detailed models far beyond theneeded level for handling simulations. They are useful to study structuralforces and displacements on the vehicle body. The vast number of parame-ters in the multi-body models also makes them inappropriate to use whenmany different vehicle configurations are used. There are also recent inter-mediate DOF handling models, such as the commercial TruckSim softwarebased on the AutoSim package. A drawback with those are that they aretargeted at engineers rather than researchers. They are very easy to use

4

when working with conventional vehicles, but have less support for uncon-ventional designs.

The present model is also targeted at future real-time simulation ap-plications. A possible use would be in hardware-in-the-loop simulations forevaluating various control equipment. This means that algebraic relationsrequiring iterative solutions during simulation have been avoided.

1.2 Related Work

In the 60s and 70s there was an interest in handling-oriented models oftractor-semitrailer vehicles. The focus was then stability, and in particu-lar jack-knifing behaviour of the vehicles. Early work was performed byEllis [Ell69, Ell88, Ell94], who presents a 4-DOF nonlinear planar dy-namic model of the tractor-semitrailer vehicle. An extension to the modelincludes roll freedom. A similar model is described in [Leu70]. Mikul-cik [Mik68, Mik71] presents a more complete 8-DOF tractor-semitrailermodel with longitudinal, lateral, heave and roll freedoms for the tractor-semitrailer, and separate yaw and pitch freedoms for the tractor and semi-trailer respectively. A Fortran implementation of this model is also pre-sented. The Fortran code includes several iterative algorithms for solvingalgebraic constraints. The model is used for analysis of jack-knifing be-haviour. AutoSim [Say92] is a tool for generating efficient simulation codefor a certain model class. The user only needs to describe the configurationof the system in terms of parameters. The equations of motion are solvedfor automatically. The major drawback with these tools in a research con-text is the closed environment they constitute. TruckSim generates modelsfor conventional vehicles. Unconventional design that might be of interestfor the research community are not included. The models may be extendedwith equations expressed in Lisp-like language, but in an environmentwhere many experimental design are to be modeled it may be better withcustom simulation models. Even though it may be a tedious work to diveinto the actual equations of a model, that may provide a lot of insight intothe model and its behaviour. A recent related work is [DYT00]. This workpresents a 8-DOF (including wheel rotations) model of a car, intended forhandling simulations similar to those targeted by the present work. Theauthors address the need for intermediate DOF models for handling stud-ies. An AutoSim model is used for validation. In [CT95, CT00] a 5-DOFtractor-semitrailer model is derived using Lagrange methods. In this modela common roll freedom for the tractor and the semitrailer is introduced. Themodel is intended for the study of lateral dynamics control in the contextof research on Automated Highway Systems. The model is evaluated withexperimental data. In [RA95] a linear 3-DOF model of a tractor-semitrailerat constant speed is derived and implemented in Simulink. The objectiveis yaw-stability analysis simulations.

2. Model Overview

The configuration of the model is illustrated in Figure 2. The model con-sists of two sprung inertial bodies, connected by the hitch. The unsprungbodies (axles) are assumed to be massless. The hitch introduces algebraicconstraints on the relative motion of the tractor and the semitrailer. These

5

algebraic relations are used to reduce the number of freedoms in the finalequations of motion. Suspension characteristics at the wheel corners are in-cluded, as well as torsional stiffness of the tractor chassis frame. A realisticmodel of the kinematics and kinetics of the hitch is included. No approxi-

Figure 2 Vehicle configuration.

mations are made when deriving the equations of motion, other than theassumption on the vehicle configuration. Approximations are introduced inthe modeling of the suspensions, axles, and tires.

The equations of motion are described in a reference system located inthe hitch, that is aligned with the unsprung tractor frame, and fixed withrespect to roll and pitch. This point on the vehicle has the unique propertyof being common for the tractor and semitrailer, which makes it possibleto reduce the equations of motion using the kinematic constraints of thehitch. This particular choice of reference system results in reasonable smallexpressions that may be handled by computer-algebra tools such as Maple.The state variables that are used to describe the motion of the vehicle arelisted in Table 1.

State variable Description

U Longitudinal velocity of the hitch

V Lateral velocity of the hitch

W Vertical velocity of the hitch

r Yaw-rate of the tractor

ψ Articulation angular velocity

φs Tractor roll angular velocity

χs Tractor pitch angular velocity

φs Semitrailer roll angular velocity

χs Semitrailer pitch angular velocity

Z Pitch vertical position

ψ Articulation angle

φ t Tractor roll angle

χ t Tractor pitch angle

φs Semitrailer roll angle

χs Semitrailer pitch angle

Table 1 State variables.

The model is modular in the sense that axle, suspension, and tire sub-

6

models are separate from the chassis dynamics. Hence it is easy to changeaxle and wheel configurations, to plug in different tire models, or to usedifferent suspension models.

The analytical mechanics background of the modeling approach is de-scribed in detail in [LU84, FC93]. In the presentation below some notationhas to be introduced. To help the reader a table of notations is included inAppendix A.

3. Vehicle configuration

The vehicle is divided into five separate parts as in Figure 3: The tractorsprung body Bts, the semitrailer sprung body Bss, the tractor unsprungbody Btu, the semitrailer unsprung body Bsu, and the hitch body ∗, or reartractor part, Bh. The main reference system is the coordinate system Stu

that is attached to the tractor unsprung mass. It is aligned with the tractor,but does not roll nor pitch. The reference system Ssu is similarly alignedwith the semitrailer. The reference systems Sts, Sss and Sh are attachedto the corresponding bodies. The reference system S∗ is the earth-fixedinertial reference system. The origins of Stu, Ssu, Sts, Sss and Sh coincideand are located at the hitch, and is denoted O.

roll, pitch

yaw

roll, pitch roll, pitch

roll

longitudinal, lateral, vertical, yawlongitudinal, lateral, vertical, yaw

PSfrag replacements

BtsBss

Btu

Bsu

Bh

StsSss

StuSsu

Sh

S∗

O

O

O

O

O

Figure 3 Vehicle parts with attached reference systems. The relative motionbetween reference systems are indicated with dashed lines with labels.

Vector variables are denoted with a bar. Vector representations with re-spect to a particular reference system is indicated by a superscript on thevector variable. Vectors are then represented as 3�1-matrices holding thecomponents of the vector with respect to the reference system indicated bythe superscript. The base vectors of the reference system Si represented in

∗Also denoted as “fifth wheel”, or “king-pin”.

7

Si are denoted by eix =

(1, 0, 0

), ei

y =(0, 1, 0

)and ei

z =(0, 0, 1

). Time deriva-

tives of a vector with respect to a certain reference system is indicated witha subscript on the differential operator as (d/dt)i qj . The shorthand nota-tion of ˙qj is used if i = j, i.e. the time derivative of qj may be expressedas the time derivative of the components of qj .

The reference systems Stu and Ssu describe yaw rotation and trans-lational longitudinal, lateral and vertical motion with respect to S∗. Theangular velocities are ω tu

tu =(0, 0, r

)T and ω susu =

(0, 0, r′

)T respectively.The velocity of O is vtu

O =(U , V , W

). The angle between Stu and Ssu is

denoted the articulation angle, and is defined as ψ =∫

rdt −∫

r′dt. Thereference system Sts describes a pitch and roll motion with respect to Stu,that is represented with the pitch angle χ t and the roll angle φ t. Corre-spondingly, Sss describes a pitch and roll motion with respect to Ssu, thatis represented by χ s and φs. The reference system Sh describes a pitch androll motion with respect to Stu, that is represented by χ t and φh. In themodeling of non-articulated vehicles it is common to express the equationsof motion in coordinate systems attached to the sprung body [KN00]. Forarticulated vehicles such as the tractor-semitrailer combination this maylead to difficulties in reducing the equations using the kinematic articula-tion constraints. For the tractor-semitrailer model the equations of motionare most conveniently expressed in the Stu and Ssu systems. Coordinatesystems are oriented according to the SAE standard, with x pointing for-ward, y to the right, and z downwards, with respect to the vehicle forwarddirection.

4. Kinematics

In the derivation of the equations of motion for the vehicle it is necessaryto have expressions for the acceleration of arbitrary points on the vehiclebody.

4.1 Motion of a point in vehicle coordinates

Denote with S∗ the earth fixed inertial reference frame, and with Si thevehicle fixed non-inertial reference frame rotating with the angular velocityω i and translating with the velocity vO, see Figure 4. For any vector q itPSfrag replacements

0

∗

P

vO

rP

vPap

ω i

Figure 4 Acceleration of a particle in a rotating and translating reference frame.

holds that(

ddt

)

∗q=

(ddt

)

iq+ ω i � q (1)

8

where ’�’ denote the cross-product operator. In particular the velocity of Pwith respect to S∗ is expressed as the sum of the translational velocity ofthe vehicle reference system, and the time derivative of the position vectorfor P, rP:

vP = vO +(

ddt

)

∗rP = vO +

(ddt

)

irP + ω i � rP = vO + ˙rP + ω i � rP (2)

The acceleration of P is computed by applying (1) on (2):

aP =(

ddt

)

∗vP =

(ddt

)

i(vO + ˙rP + ω i � rP) + ω i � (vO + ˙rP + ω i � rP)

= ˙vO + ω i � vO + ˙ω i � rP + ω i � (ω i � rP) + 2ω i � ˙rP + ¨rP (3)

4.2 Coordinate transformations

Let qts be a vector representation in the tractor system Sts. Then qtu =Rtu

ts qts is the representation of this vector in Stu. The rotational transfor-mation Rtu

ts is defined as consecutive pitch and roll transformations with

Rtuts =

1 0 0

0 cosφ t − sinφ t

0 sinφ t cosφ t

cos χ t 0 sin χ t

0 1 0

− sin χ t 0 cos χ t

(4)

Correspondingly a vector qss in the semitrailer system Sss is representedin Ssu as qsu = Rsu

ss qss with

Rsuss =

1 0 0

0 cosφs − sinφs

0 sinφs cosφs

cos χs 0 sin χs

0 1 0

− sin χs 0 cos χs

(5)

Transformations between the semitrailer system Ssu and the tractor systemStu is described with qtu = Rtu

su qsu with

Rtusu =

cosψ sinψ 0

− sinψ cosψ 0

0 0 1

(6)

The hitch kinematics is such that the roll angle φh of the tractor rearpart Bh is given by

φh = φs cosψ + χs sinψ (7)

Thus it is equal to the roll angle of the semitrailer at zero articulation angle,and equal to the semitrailer pitch angle at 90 degrees articulation angle.The tractor rear part Bh always has the same pitch angle as the tractorfront part Bts. Transformations between the tractor rear part system Shand the tractor system Stu is described with qtu = Rtu

h qh with

Rtuh =

1 0 0

0 cosφh − sinφh

0 sinφh cosφh

cos χ t 0 sin χ t

0 1 0

− sin χ t 0 cos χ t

(8)

9

4.3 Motion of a point on the tractor body

Let rtsP denote the position of an arbitrary point P on the tractor body B ts,

expressed in the tractor fixed reference system Sts. Then

rtuP = Rtu

ts rtsP (9)

The velocity of P is then

vtuP = vtu

O +(

ddt

)

turtu

P + ω tutu � rtu

P = vtuO + R

tuts rts

P + ω tutu � Rtu

ts rtsP (10)

and the acceleration

atuP =

(ddt

)

tuvtu

P + ω tutu � vtu

P = ˙vtuO + ω tu

tu � vtuO + R

tuts rts

P + ˙ω tutu � Rtu

ts rtsP

+ 2ω tutu � R

tuts rts

P + ω tutu �

(ω tu

tu � Rtuts rts

P

)(11)

4.4 Motion of a point on the semitrailer body

Let rssP denote the position of an arbitrary point P on the semitrailer body

Bss, expressed in the semitrailer fixed reference system Sss. Then

rsuP = Rsu

ss rssP (12)


vsuP = vsu

O +(

ddt

)

sursu

P + ω susu � rsu

P = vsuO + R

suss rss

P + ω susu � Rsu

ss rssP (13)

and the acceleration

asuP =

(ddt

)

suvsu

P + ω susu � vsu

P = ˙vsuO + ω su

su � vsuO + R

suss rss

P + ˙ω susu � Rsu

ss rssP

+ 2ω susu � R

suss rss

P + ω susu � (ω su

su � Rsuss rss

P ) (14)

4.5 Motion of a point on the hitch body

Let rhP denote the position of an arbitrary point P on the hitch body Bh,

expressed in the rear tractor fixed reference system Sh. Then

rtuP = Rtu

h rhP (15)


vtuP = vtu

O +(

ddt

)

turtu

P + ω tutu � rtu

P = vtuO + R

tuh rh

P + ω tutu � Rtu

h rhP (16)

5. Kinetics

The free-body diagram of Figure 5 introduces the forces and moments thatare used in the kinematic analysis. The bodies B ts and Bss carry the masses

10

PS

frag

repl

acem

ents

Bts

Bss

Btu

Bsu

Bh

Sts

Sss

Stu

Ssu

Sh

S∗

O

O

O

OOm

tsk −

Fs,

fl

−F

s,fr−M

s,fl

−M

s,fr

−M′ ts

−F′ ts

Fs,

fl

Fs,

fr

Ms,

fl

Ms,

fr

mssk

−F

s,sl

−F

s,sr−M

s,sl

−M

s,sr

−M′ ss

−F′ ss

Fs,

sl

Fs,

sr

Ms,

sl

Ms,

sr

−F

s,rl

−F

s,rr

−M

s,rl

−M

s,rr

M′ ts+

M′ ss

F′ ts+

F′ ss

Fs,

rl

Fs,

rr

Ms,

rl

Ms,

rr

Fw

,fl

Fw

,fr

Fw

,rl

Fw

,rr

Fw

,sl

Fw

,sr

Figure 5 Vehicle free-body diagram.

mts and mss, with the inertial tensors I ts =∫

Bts

(r2

P1− rP rTP

)dmP and Iss =∫

Bss

(r2

P1− rP rTP

)dmP with respect to O.

The fundamental equations of dynamics in Euler’s formulation postu-lates that

F �∫

B

¯dFP =∫

BaPdmP (17a)

MO �∫

Brp � ¯dFP =

∫

BrP � aPdmP (17b)

11

Thus for the tractor

Fts − F′ts =∫

Bts

aPdmP (18a)

Mts − M ′ts =

∫

Bts

rP � aPdmP (18b)

where Fts and Mts are the sum of external moments and forces acting onBts, and F′ts and M ′

ts are the internal reaction forces and moments fromBh. Accordingly for the semitrailer

Fss − F′ss =∫

Bss

aPdmP (19a)

Mss − M ′ss =

∫

Bss

rP � aPdmP (19b)

The static kinetic constraints that arise from the massless free body Bh,are

F′ts + F′ss + Fh = 0 (20a)M ′

ts + M ′ss + Mh = 0 (20b)

where Fh and Mh are the external forces and moments acting on Bh.

5.1 Elimination of internal forces and moments

The internal forces F′ts and F′ss may be eliminated by combining (18a),(19a) (20a), yielding

Fts + Fss + Fh =∫

Bts

aPdmP +∫

Bss

aPdmP (21)

or

Ftuts + Rtu

su Fsuss + Ftu

h =∫

Bts

atuP dmP + Rtu

su

∫

Bss

asuP dmP (22)

when represented in Stu. The integrals evaluate to terms including compo-nents of the center of gravity locations rts

CM ,ts and rssCM ,ss, and the masses

mts and mss.A little more care has to be taken when eliminating internal moments,

because of the torque transfer characteristics of the hitch. Regard that thehitch transfer moment to the semitrailer along the ess

x axis at 0 articulationangle, and along the ess

y axis at 90 degrees articulation angle:

mssss = mss

ss,x y ( cosψ , sinψ , 0 )T = mssss,x y cosψ ess

x + mssss,x y sinψ ess

y (23)

The moment transfer to the tractor is in all etsx , ets

y and etsz directions, and

is determined by the roll stiffness in the etsx direction:

mtsts = ( Cc (φ t − φh) , mts

ts,y, mtsts,z )T = Cc (φ t − φh) ets

x + mtsts,y ets

y +mtsts,z ets

z

(24)

12

Now (20b) may be expressed in B tu as

Rtuts

(Cc (φ t − φh) ets

x +mtsts,y ets

y +mtsts,z ets

z

)

+ Rtuss

(mss

ss,x y cosψ essx + mss

ss,x y sinψ essy

)+ Rtu

h Mhh = 0 (25)

or

( Rtuss

(cosψ ess

x + sinψ essy

)Rtu

ts essy Rtu

ss essz )

mssss,x y

mtsts,y

mtsts,z

= −Rtuh Mh

h − Cc (φ t − φh) Rtuts ets

x (26)

which can be solved for mssss,x y, mts

ts,y and mtsts,z. Thus (18b) and (19b) may

be represented as

M tuts − Rtu

ts mtsts =

∫

Bts

rtuP � atu

P dmP (27)

M suss − Rsu

ss mssss =

∫

Bss

rsuP � asu

P dmP (28)

The inertial tensors I ts and Iss expressed in the Sts and Sss systems givesthe inertial matrices

I tsts =

∫

Bts

(rts

P21− rts

P rtsP

T)dmP

=

∫Bts(yts

P )2 + (ztsP)2dmP −

∫Bts

xtsP yts

P dmP −∫

Btsxts

P ztsP dmP

−∫

Btsyts

P xtsP dmP

∫Bts(zts

P )2 + (xtsP )2dmP −

∫Bts

ytsP zts

P dmP

−∫

Btszts

P xtsP dmP −

∫Bts

ztsP yts

P dmP∫

Bts(xts

P )2 + (ytsP)2dmP

(29)

and

Issss =

∫

Bss

(rss

P21− rss

P rssP

T)dmP

=

∫Bss(yss

P )2 + (zssP )2dmP −

∫Bss

xssP yss

P dmP −∫

Bssxss

P zssP dmP

−∫

Bssyss

P xssP dmP

∫Bss(zss

P )2 + (xssP )2dmP −

∫Bss

yssP zss

P dmP

−∫

Bsszss

P xssP dmP −

∫Bss

zssP yss

P dmP∫

Bss(xss

P )2 + (yssP )2dmP

(30)

The integrals in (27) and (28) evaluate to terms with elements from theseinertial matrices.

Now (22), (27) and (28) forms the complete set of equations of motionfor the tractor-semitrailer combination vehicle. A following section providesa more detailed description of the external forces that appear in theseequations.

13

5.2 State equations

Equations (22), (27) and (28) may be written in matrix form as

Ftu1 + Rtu

su Fsuss + Ftu

h

M tuts − Rtu

ts M ′ tutu

M suss − Rsu

ss M ′ssss

=

∫Bts

atuP dmP + Rtu

su

∫Bss

asuP dmP∫

Btsrtu

P � atuP dmP∫

Bssrsu

P � asuP dmP

(31)

Introduce the state vector

ξ = (ξ1, ξ2) (32a)

with

ξ1 = (U , V , W, r,ψ , φ t, χ t, φs, χs) (32b)

and

ξ2 = (Z,ψ , φ t, χ t, φs, χs) (32c)

The right hand side of (31) may be rewritten such that

Ftuts + Rtu

su Fsuss + Ftu

h

M tuts − Rtu

ts mtsts

M suss − Rsu

ss mssss

= H(ξ )dξ1

dt− F(ξ ) (33)

The complete state equations are

dξ1

dt= H(ξ )−1

F(ξ ) +

Ftuts + Rtu

su Fsuss + Ftu

h

M tuts − Rtu

ts mtsts

M suss − Rsu

ss mssss

(34)

dξ2

dt= Eξ1 (35)

with

E =

0 0 1 0 0 0 0 0 0

0 0 0 1 0 0 0 0 0

0 0 0 0 0 1 0 0 0

0 0 0 0 0 0 1 0 0

0 0 0 0 0 0 0 1 0

0 0 0 0 0 0 0 0 1

(36)

The elements of the matrix H(ξ ) and the vector F(ξ ) are large expressionsthat need to be computed with computer algebra software such as Maple.The matrix H(ξ ) is not possible to invert analytically, since it would leadto very large expressions. Instead this matrix is computed numerically ateach integration step, and inverted numerically.

The large expressions in the elements of H(ξ ) and F(ξ ) result from thecross-products of the position and acceleration vectors of Sections 4.3–4.4.The acceleration vectors also include first and second time derivatives of

14

the transformation matrices in Section 4.2. As an illustration of the sizeof the analytical expressions in the matrices, the number of terms in eachelement is given below.

H :

3 1 1 5 2 1 2 2 2

1 3 1 4 2 3 2 2 2

1 1 2 1 1 3 2 3 2

1 4 3 8 1 11 6 1 1

4 1 2 11 1 6 8 1 1

3 2 1 9 1 6 6 1 1

4 4 3 8 8 1 1 11 6

4 4 2 11 11 1 1 6 8

5 5 1 9 9 1 1 6 6

F :

12

15

14

41

42

34

74

79

57

(37)

The terms are nonlinear expressions of the state variables.

5.3 External forces

The external forces that act on the vehicle are gravitational forces, tyre-road contact forces, and wind-drag forces. Wind-drag forces are propor-tional to the squared resultant wind velocity [KN00].

Fwind,x

Fwind,y

Fwind,z

=

Caer,x Axρ2 U2

wind,x

Caer,yAyρ2 U2

wind,y0

(38)

where Caer,x/y are aerodynamic drag coefficients, Ax/y front and side ef-fective areas, and Uwind,x/y the resultant wind velocity. The gravitationalforces are

Fk,ts =∫

Bts

kdmP = mtsk etuz (39)

Fk,ss =∫

Bss

kdmP = mssk esuz (40)

The tyre-road contact forces are obtained from the tire model, which isdescribed in a following section. These forces are transfered to the sprungchassis through the suspension, and depend on suspension characteristicsand axle configurations. The suspension system determines the verticalforces as well as pitch and roll moments.

The axle and suspension kinetics are derived from analyzing the free-body diagram of Figure 6. The indices “f ”, “r” and “s” on the variables inthe diagram denote front axle, rear axle and semitrailer axle. (There maybe several of each sort.) The indices “l” and “r” denote the left or right sideof axle†. The description of the axle kinetics is presented for an arbitraryaxle. The axles kinetics is described in the Stu and Ssu for the tractor andsemitrailer axles respectively. Axle indices are therefore dropped in thefollowing.

†Even if the same index is used for “rear axle” and “right side”, the meaning is alwaysclear from the context.

15

PSfrag replacements

Fs,l,x

Fs,l,x

Fs,r,x

Fs,r,x

Fs,l,y

Fs,l,y

Fs,l,z

Fs,l,z

Fs,r,y

Fs,r,y

Fs,r,z

Fs,r,zMs,x

Fw,l,x

Fw,l,x

Fw,r,x

Fw,r,x

Fw,l,y

Fw,l,y

Fw,l,z

Fw,l,z

Fw,r,y

Fw,r,y

Fw,r,z

Fw,r,z

rs,l,y

rs,l,y

rs,r,y

rs,r,y

rw,l,y

rw,l,y

rw,r,y

rw,r,y

Ms,y

Stu/Ssu

Stu/Ssu

Stu/Ssu

eZ + rs,l,ze

eZ + rs,l,ze

eZ + rs,r,ze

eZ + rs,r,ze

etux /esu

x

etux /esu

x

etuy /esu

y

etuy /esu

y

etuz /esu

z

etuz /esu

z

Figure 6 Tractor front axle and suspension kinetics. Tractor rear axle and semi-trailer axle have corresponding configurations.

Denote the suspension forces at corner i ∈ {l, r} by Fs,i, the suspensionmoments by Ms,i, and the wheel forces by Fw,i. Now Fs,i,z, Fw,i,x, Fw,i,y,Ms,i,x, Ms,i,z may be expressed directly in terms of position and velocity ofthe suspension corner, which can be computed from the state variables,while Fs,i,x, Fs,i,y, Fw,i,z, Ms,i,y need to be solved for. Since the axle and thebody are two rigid bodies that have two supporting connections aligned inthe etu

y , esuy directions, there are no internal moments acting in the etu

x , esux

and etuz , esu

z directions. With an additional roll-stiffener an extra momentin the etu

x , esux direction may be added. Let Ms,x = Ms,l,x + Ms,r,x and Ms,y =

Ms,l,y + Ms,r,y denote the lumped suspension moments on one axle. Thisreduces the number of free variables.

16

5.4 Axle balance equations

The tyre forces Fw,l,x, Fw,l,y, Fw,r,x, Fw,r,y are assumed to be expressed asFw,l,x = Fw,l,zµ l,x, Fw,l,y = Fw,l,zµ l,y, Fw,r,x = Fw,r,zµr,x , Fw,r,y = Fw,r,zµr,y,where µ l,x, µ l,y, µr,x , µr,y are coefficients of friction given from the tyremodel. This assumption is commented further at the end of this section.

Let ∆rs,l,z, ∆rs,r,z, vs,l,z and vs,r,z denote the suspension travel and ve-locity at the left and right suspension corners at a given time. These vari-ables may be computed from the state variables using the transformationsof Section 4. The vertical suspension forces may now be computed usingany desired mapping. In this presentation we introduce the simple linearsuspension

Fs,l,z = −Cs∆rs,l,z − Dsvs,l,z (41a)Fs,r,z = −Cs∆rs,r,z − Dsvs,r,z (41b)

In the implementation of the model presented in this report nonlinearasymmetrical dampers based on lookup-tables are used.

The known suspension moments are

Ms,x = −Cr(∆rs,r,z − ∆rs,l,z) (42a)Ms,z = 0 (42b)

where an anti-roll bar with stiffness Cr is introduced to provide additionalroll stiffness.

Balance of forces and moments result in the following set of equations:

Fs,l,x + Fs,r,x + Fw,l,x + Fw,r,x = 0 (43a)Fs,l,y + Fs,r,y + Fw,l,y + Fw,r,y = 0 (43b)Fs,l,z + Fs,r,z + Fw,l,z+ Fw,r,z = 0 (43c)

−rs,l,z Fs,l,y − rs,r,z Fs,r,y + rs,l,y Fs,l,z + rs,r,y Fs,r,z+rw,l,y Fw,l,z+ rw,r,y Fw,r,z + Ms,x = 0 (43d)

rs,l,z Fs,l,x + rs,r,z Fs,r,x + Ms,y = 0 (43e)−rs,l,y Fs,l,x − rs,r,y Fs,r,x − rw,l,y Fw,l,x − rw,r,y Fw,r,x = 0 (43f)

Recall that the wheel forces Fw,l,y and Fw,l,x are assumed to depend linearlyon the normal forces. Hence, the equations (43a–43f) can be rewritten as

Fs,l,x + Fs,r,x − µ l,x Fw,l,z− µr,x Fw,r,z = 0 (44a)Fs,l,y + Fs,r,y − µ l,y Fw,l,z− µr,y Fw,r,z = 0 (44b)

Fw,l,z+ Fw,r,z = −Fs,l,z − Fs,r,z (44c)−rs,l,z Fs,l,y − rs,r,z Fs,r,y + rw,l,y Fw,l,z+ rw,r,y Fw,r,z = (44d)

−Ms,x − rs,l,y Fs,l,z − rs,r,y Fs,r,z

rs,l,z Fs,l,x + rs,r,z Fs,r,x + Ms,y = 0 (44e)rs,l,y Fs,l,x + rs,r,y Fs,r,x − rw,l,yµ l,x Fw,l,z− rw,r,yµr,x Fw,r,z = 0 (44f)

In addition, it is assumed that side forces are equally carried by the sus-pension

Fs,l,y = Fs,r,y (45)

17

Equations (44a–44f) together with (45) give

1 1 0 0 µ l,x µr,x 0

0 0 1 1 µ l,y µr,y 0

0 0 0 0 1 1 0

0 0 −rs,l,z −rs,r,z rw,l,y rw,r,y 0

rs,l,z rs,r,z 0 0 0 0 1

rs,l,y rs,r,y 0 0 rw,l,yµ l,x rw,r,yµr,x 0

0 0 1 −1 0 0 0

⋅

Fs,l,x

Fs,r,x

Fs,l,y

Fs,r,y

Fw,l,z

Fw,r,z

Ms,y

=

0

0

−Fs,l,z − Fs,r,z


0

0

0

(46)

Roll-over Conditions During heavy cornering the right or left wheelmay not stay in contact with the ground. The wheel normal force is thenzero, i.e. Fw,l,z = 0 or Fw,r,z = 0, and the solution of the force and momentbalance in Equation (46) is not valid. To still achieve at least approxi-mately correct behaviour from the model under such conditions a simplisticsolution is presented in this section. The basic idea is to replace the freevariable (Fw,l,z or Fw,r,z) that disappears from the balance equation withanother. For this purpose the axle roll angle ϕ is introduced as a new de-gree of freedom. Two simplifying assumptions are also introduced to avoida nonlinear axle balance equations that might require iterative solutionmethods: (i) It is assumed that spring characteristics are linear. (ii) Thedamper force does not depend on the roll velocity of the suspension axle ϕ .

The case with Fw,l,z = 0, leading to ϕ > 0, is described in the following.(The case with Fw,r,z = 0 and ϕ < 0 is treated analogously.) The verticalsuspension forces in (41a) and (41b) then become

Fs,l,z = −Cs(∆rs,l,z − (rw,r,y − rs,l,y) sin(ϕ )

)− Dsvs,l,z

= −Cs∆rs,l,z − Dsvs,l,z+ Cs(rw,r,y − rs,l,y) sin(ϕ )� Fs,l,z + kl sin(ϕ ) (47a)

Fs,r,z = −Cs (∆rs,r,z − (rw,r,y − rs,r,y) sin(ϕ )) − Dsvs,r,z

= −Cs∆rs,r,z − Dsvs,r,z + Cs(rw,r,y − rs,r,y) sin(ϕ )� Fs,r,z + kr sin(ϕ ) (47b)

By applying (47a) and (47b) to (44a)–(44f) the following set of linear equa-tions is obtained:

18

1 1 0 0 0 −µr,x 0

0 0 1 1 0 −µr,y 0

0 0 0 0 kl + kr 1 0

0 0 −rs,l,z −rs,r,z rs,l,ykl + rs,r,y kr rw,r,y 0

rs,l,z rs,r,z 0 0 0 0 1

rs,l,y rs,r,y 0 0 0 −rw,r,yµr,x 0

0 0 1 −1 0 0 0

⋅

Fs,l,x

Fs,r,x

Fs,l,y

Fs,r,y

sin(ϕ )Fw,r,z

Ms,y

=

0

0

−Fs,l,z − Fs,r,z


0

0

0

(48)

Tires with Nonlinear Normal Force Dependence If the assump-tions on linear normal force dependence of the tire forces Fw,x = Fw,zµ x

and Fw,y = Fw,zµ y do not hold, but instead the tire forces are describedby general nonlinear functions (Fw,z, ξ ) =→ Fw,x and (Fw,z, ξ ) =→ Fw,y, thenthis would lead to nonlinear algebraic relations. To compute the tire lat-eral forces one would need the tire vertical forces, and to compute the tirevertical forces, the lateral forces would be needed. This would need to behandled with iterative solution methods, and would increase the computa-tional complexity. A way to handle this is to introduce a piecewise linearapproximation as

Fw,x = mi + kiFw,zµ x, with i such that Fw,z ∈ F i (49)

A set of triples {(F i, mi, ki)}, i ∈ 1, . . . n where F i is the interval of thenormal force where the linear approximation is valid, is defined. Then theaxle balance equations are solved for each i, and the solution for which thesolved normal forces Fw,l,z, Fw,r,z belongs the assumed interval F i is cho-sen. This means solving a finite number (n) of axle balance equations andchoosing one solution. The advantage is that this computation has a knowncomputation time, which is not necessarily true for iterative methods. Of-ten it is possible to find reasonably accurate piece-wise linear models ofthe normal load dependence.

6. Tire Models

The longitudinal tire slip λ is defined as

λ = ω Rw − uu

(50)

where u, v are the longitudinal and lateral velocities of the wheel centerexpressed in a reference system aligned with the wheel, ω is the wheel

19

angular velocity, and Rw is the effective wheel radius. The lateral side-slipα is defined as

tan α = − vu

(51)

Many tire models are formulated as mappings from λ , α and the tire ver-tical load Fw,z to the longitudinal and lateral adhesion forces Fw,x and Fw,y.The adhesion coefficients µ x and µ y are defined as

µ x =Fw,x

Fw,z(52a)

µ y =Fw,y

Fw,z(52b)

The adhesion force characteristics for a real tire at pure cornering andpure braking are shown in Figure 7. At combined braking and cornering

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

λ

X/Z

0 0.2 0.4 0.6 0.8 10

0.2

0.4

0.6

0.8

1

α

Y/Z

Figure 7 Experimental tire adhesion coefficients at pure braking and pure cor-nering. The last data value for the lateral adhesion coefficient is estimated as beingequal to the longitudinal adhesion coefficient for λ = 1. The experimental data isobtained from Volvo Truck Corporation [Edl91]† .

the adhesion coefficients are reduced compared to pure braking and purecornering. This is illustrated for a real tire in Figure 8. The choice oftire model depends on in which region the model is supposed to operate.For maneuvers with pure braking or pure cornering and small slips thetires behave approximately linearly and a simple tire model may suffice.At large slips or combined braking and cornering it may be necessary touse more sophisticated models. Two different models are presented in thefollowing sections.

6.1 Linear Tire Model

The simplest tire model is the linear model

Fw,x = Cλλ (53a)Fw,y = Cαα (53b)

The model (53) has the drawbacks of not including limitations of avail-able tire adhesion forces, and not modeling the coupling between longitu-dinal and lateral adhesion forces at combine cornering and braking. It maystill be useful for modeling tire behaviour in the small-slip region.

20

−1 −0.8 −0.6 −0.4 −0.2 0−1

−0.8

−0.6

−0.4

−0.2

0

Fx/Fz

Fy/

Fz

α = 4.7°α = 9.8°

Figure 8 Experimental tire adhesion coefficients at combined braking and cor-nering for two fixed α and varying λ . The experimental data is obtained from VolvoTruck Corporation [Edl91]† .

6.2 Slip Circle Model

The slip circle model [SPP96] is a generic tyre model for combined brakingand cornering based on models of pure braking and cornering. Introducethe dimensionless slip variable s

s =√

λ2 + sin2 α (54)

(It is common to describe the side-slip with the dimensionless entity sinαinstead of α . For small α the difference is negligible.) Define the slip angleβ as

tan β = sinαλ

(55)

The tyre force is assumed to be counter-directed to the slip vector describedby s and β .

Now introduce the pure cornering and braking mappings from slips totyre forces f : (λ , Fw,z) =→ Fw,x and k : (α , Fw,z) =→ Fw,y. In the slip-circlemodel the magnitude of the combined tyre force is now described as

F = f (s) cos2 β + k(s) sin2 β (56)

Now

Fw,x = F cos β (57a)Fw,y = F sin β (57b)

21

It is seen that pure cornering and braking are restored for β = 0 andβ = π/2. For other β the combined tyre forces lies on a curve that is“close” to an ellipse. The slip circle model is compared with experiments inFigure 9. The mappings f and k may be chosen as experimentally acquired

−1 −0.5 0−1

−0.8

−0.6

−0.4

−0.2

0

Fx/Fz

Fy/

Fz

α 4.7°α 9.8°α 4.7° (model)α 9.8° (model)

Figure 9 Slip circle model and experimental adhesion coefficients at combinedbraking and cornering for two fixed α and varying λ . The experimental data isobtained from Volvo Truck Corporation [Edl91]† .

lookup-tables.

7. Wheel Dynamics

The rotational dynamics of the wheels are modeled as

Jwddt

ω = Mω − RwFw,x (58)

where Jw is the moment of inertia of the wheel, and Mω the driving orbraking moment.

In many cases the wheel rotational dynamics are not interesting tomodel, and the longitudinal slips λ may be used directly as inputs to thevehicle model.

8. Implementation

The general procedure to realize the dynamics described in Section 5 is toanalytically compute the matrices F(ξ ) and H(ξ ) using computer algebrasoftware. An automated tool to translate these expressions to a computerlanguage supported by the simulation software is then used. In this workMaple was used to compute F(ξ ) and H(ξ ). Maple has support for trans-lation of expressions to the programming language C. The expressions forF(ξ ) resulted in around 300 lines of C-code, and the expressions for H(ξ )in around 200 lines. The generated C-files were included in Matlab mex-file

†The details on the tire experiments are classified corporate information.

22

functions defined to return the matrices as functions of the state ξ . The po-sition, velocity and acceleration relations of Sections 4.3–4.5 was similarlycomputed in Maple and translated to mex-file functions. These relationsare needed to compute suspension and wheel motion that are used for theaxle balance equations of Section 5.4, and the tire models of Section 6.The resulting simulation code is modular in the sense that the dynamicequations, the suspensions and the axle load balance equations, and thetire model equations are separated. This makes it easy to substitute one ofthese sub-models for another. The organization of the resulting Simulinkmodel is illustrated in Appendix B. In the present implementation the chas-sis dynamics, suspension and axle load balance equations, and wheel/tiremodels are implemented as separate S-functions in Simulink. This is notthe optimal implementation structure for simulation-speed performance,but results in a well organized code that is easy to overview.

The implemented vehicle configuration is a 4x2 tractor-semitrailer. Thesemitrailer has three axles. The model includes wheel rotational dynamicswith torque inputs, but may also be used without wheel dynamics withlongitudinal slips as direct inputs. The wheel torques Mw,i or the longitu-dinal slips λ i with i ∈ { f l, f r, rl, rr, s1l, s1r, s2l, s2r, s3l, s3r}, and the frontwheel steering angle δ are the inputs to the model. Wind-drag forces arenot included in the implementation. The parameters of the model are listedin Table 2.

Variable road-surface conditions are easily modeled by providing anadhesion reduction factor as a lookup-table in position xy-coordinates.

The mass distribution of the tractor and semitrailer sprung bodies weremodeled as combinations of rectangular boxes with homogeneous density.This makes it ease to arrive at desired inertia properties. See Appendix C.

A 3D-animation routine has been developed to illustrate simulationresults. This routine is based on standard Matlab graphics, and produceoutput as shown in Figure 10.

8.1 Real-Time Performance

Application of the present model in real-time simulation has been consid-ered. For this reason all nonlinear algebraic equations have been avoided.Such equations would otherwise result in algebraic loops that require itera-tive solution algorithms. With such algorithms it is difficult to obtain guar-anteed convergence time, which makes it difficult to guarantee real-timeperformance. The present model have a deterministic number of floating-point operations at each simulation step (for fixed time-step integrationmethods). This prerequisite for a real-time simulation model is thereforefulfilled.

The present implementation does not simulate in real-time. The sim-ulation speed is around on tenth of real-time for variable-step Simulinksolver ode45 Dormand-Prince on a standard Pentium II PC running at 366MHz. There are several possible reasons for this:

1. The number of floating point operations is too large in each step.

2. The implementation code-structure is inefficient.

3. The model contains stiff modes forcing the solver to use small steps.

23

Chassis

mts Tractor sprung mass

mss Semitrailer sprung mass

rCG,ts Tractor sprung body CM location

rCG,ss Semitrailer sprung body CM location

I ts Tractor sprung body inertia matrix

Iss Semitrailer sprung body inertia matrix

Cc Tractor frame torsional stiffness

Axles and suspensions∗

rw,l Left wheel location

rw,r Right wheel location

rs,l Left suspension location

rs,r Right suspension location

Cs Suspension vertical stiffness

Ds Suspension vertical damping coefficient

Cr Roll-bar stiffness

hs Unloaded suspension length.

Wheel and tires∗

Rw Wheel radius

Cλ† Tire longitudinal stiffness

Cα† Tire cornering stiffness

f : λ =→ µ x‡ Longitudinal adhesion coefficient mapping at pure

braking (lookup-table)k : α =→ µ y

‡ Lateral adhesion coefficient mapping at pure corner-ing (lookup-table)

Table 2 Model parameters.

The first reason is less likely. Even if there are many floating-point opera-tions to compute in the model, those are very fast to execute. Matlab hasvery efficient numerics using BLAS/LAPACK. The second reason is defini-tively an issue. The code has been structured for user simplicity rather thanoptimal performance. In particular the partitioning of the model on manyS-functions is inefficient. Collecting everything in one S-function wouldbring down the number of S-function calls by a magnitude. The possibilityto implement this S-function in C should probably also be considered. Thethird reason in the list may also be an issue. There are indications that theparameters used in the validation simulations lead to fast modes. Duringtransient behaviour the variable-step solver bring down the step-length tovery small values. Efforts are needed to eliminate any stiff behaviour.

24

9. Validation

Experimental data for a 4x2 tractor-semitrailer vehicle recorded on a test-track was kindly provided by Volvo Truck Corporation, together with somedata on the vehicle‡. The experiments were designed for other purposesthan the validation of the present model, and were performed before thework on this model started. A set of model parameters corresponding tothe tested vehicle was derived. Since the provided data did not cover thecomplete set of parameters of the model, some assumptions were made. Inparticular it is difficult to determine tire data and suspension data withhigh accuracy. The provided tire data did not give enough information forthe slip-circle model. Therefore a linear tire model was used in the valida-tion. No information except from approximate mass and dimensions wasprovided on the semitrailer parameters. Nonlinear lookup-table mappingsfor the damper characteristics were provided, and was included in themodel. Most of the 31 test-scenarios only included steering actions. Thesteering actions were of moderate magnitude, in the sense that they onlygenerate small lateral slips. Therefore the linear tyre model used in thevalidation simulations is probably quite accurate. Some scenarios seem toinclude braking actions in combination with the steering. Unfortunately,no information on how the braking was applied was provided, which makesit difficult to use the data for validation.

The recorded variables of interest were: front wheel angles, vehiclespeed, kingpin-angle, yaw-rate, and suspension travel for the four cornersof the tractor. The average of the left and right front wheels angles wereused as inputs to the model.

The linear constant speed 4-DOF dynamic tractor-semitrailer modelderived in [Gäf01] was tuned with parameters corresponding to those ofthe 9-DOF model. This 4-DOF model has the lateral velocity V , yaw-rater, articulation angular rate ψ , and articulation angle ψ as state variables.

Both the 4-DOF and the 9-DOF models were simulated with the averagefront steering angle of the experimental data as inputs. Results from achoice of scenarios are presented in Appendix D. The 4-DOF model onlyreproduces the kingpin-angle and the yaw-rate.

9.1 Lane-Change Maneuver

Appendices D.1 and D.2 shows the typical results from two lane-changemaneuvers. Both models show good accordance with experimental datawith respect to kingpin-angle and yaw-rate. The kingpin-angle peaks area bit too large on the models. The suspension travel of the 9-DOF modelis well reproduced on the rear axle, but significantly smaller on the frontaxle.

9.2 Step-Steering Maneuver

Appendices D.3 and D.4 shows the typical results from two step-steeringmaneuvers. The results are similar to those of the lane-change maneuvers.The models performs well. An overshoot in the articulation angle is present

‡The detailed data on the vehicle tests are classified corporate information.∗One set of parameters for each instance of the object class.†For linear tire model.‡For slip-circle tire model.

25

in the model results for the first maneuver, but absent in the second. Thereason for this is difficult to understand without more information on theexperimental setup.

9.3 Random Steering Maneuvers

Appendix D.5 shows the results from random-steering maneuvers. The re-sults are similar to those of the previous maneuvers. One can note that thelongitudinal speed decreases slightly during the maneuvers for the 9-DOFmodel. This is because the experiments were performed under cruising con-ditions, while the model simulates a free-rolling scenario. The energy lossin the front tires during steering then reduces the speed. The overshootsin the kingpin-angle are still present. There are also some discrepancies inthe reproduction of the yaw-rate in this scenario. The reproduction of therear axle suspension travel is good.

9.4 Comments on the Results

Some re-tuning of parameters were performed to improve the validationresults. Still several of the parameters are uncertain, and would requiremore accurate values to obtain optimal results. It was noted that the tireparameters have large impact on the reproduction of the king-pin angleand the yaw-rate. The discrepancies in the front suspension travel maypartly be explained by uncertainty in the position of the suspension travelsensors.

The motivation for using the 9-DOF model instead of the simpler 4-DOF model is the inclusion of the tire normal forces. Those forces werenot measured in the experiments, but they correspond to the suspensiontravel. Since the suspension travel is reasonable well reproduced by the9-DOF model it is concluded that the tire normal forces are equally wellreproduced. In the present scenarios the linear model performs equallywell compared to the 9-DOF model with respect to kingpin-angle and yaw-rate. Hence good reproduction of normal forces are not needed to reproducethe planar behaviour of these scenarios. Still it expected that the loadtransfer will have significant influence in the case of unilateral brakingactions, which may be applied by yaw-stabilization systems. To really vali-date the benefits of good normal-force modeling more aggressive scenariosare needed, where the load-transfer is significant and the tires operate inthe nonlinear region. Experiments with combined braking and corneringwould probably also require more accurate models because of additionalload-transfer. Experiments with “large” state-variable registerings, wherethe linear model is less accurate, would also be interesting to investigateto evaluate possible advantages of using the nonlinear model. This couldalso include experiments on slippery surfaces with little adhesion.

To conclude, the validation results seem reasonable considering the un-certainties in the parameters. Qualitatively the performance seems realis-tic.

10. Simulations

In appendix E the output from two lane-change maneuvers are presented.In these simulations the slip-circle model is used. The first simulation

26

Figure 10 Matlab 3D-animation of the vehicle.

shows a free-rolling violent lane-change maneuver. Wheel normal forcesare just above zero in the critical part of the maneuver. In the second ma-neuver the tractor rear wheel are locked with full braking. Not surprisinglythis leads to a jack-knifing accident. Note that some tire normal forces be-come zero at certain time intervals. The method described in Section 5.4for roll-over conditions is not used in these simulations. The normal forcesare simply limited to be positive. This will of course reduce the validity ofthe model when the limits are hit. This simulation is intended to show aclearly nonlinear behaviour of the model. The plots are merely examplesof some of the outputs that may be extracted from the simulations.

11. Future Work

Work is initiated on including unsprung masses in the equations of motion.Real-time optimizations in line with the discussion of Section 8.1 should bedone to speed up simulations. Further validation is needed to investigatethe ability of the model to reproduce correct nonlinear behaviour.

12. Conclusions

A nonlinear dynamic 9-DOF tractor-semitrailer model has been presented.Validation of the model indicates that load-transfer is accurately modeled.This implies that the model may give realistic results in simulation of han-dling maneuvers near and beyond the adhesion limits. Hence the model canbe suitable for studies on advanced chassis control in handling-maneuvers.

27

13. Acknowledgements

This work is part of the DICOSMOS (DIstributed COntrol of Safety criticalMOtion Systems!) project, funded by NUTEK under the Complex Techni-cal Systems program, with project nr. P11762-2, and by Volvo TechnologicalDevelopment (VTD). Mats Andersson (VTD) has been an initiator of thework on the semi-trailer combination vehicle, and has been a valuablesource of information and ideas. Niklas Fröjd at Volvo Truck Corporationhas been very helpful with providing validation data, and general com-ments on modeling issues.

28

14. References

[CGS00] V. Claesson, M. Gäfvert, and M. Sanfridson. Proposal for a dis-tributed computer control system in a heavy-duty truck. Tech-nical Report No. 00-16, Department of Computer Engineering,Chalmers Univ. of Technology, 2000. DICOSMOS Report.

[CT95] Chieh Chen and Masayoshi Tomizuka. Dynamic modeling of ar-ticulated vehicles for automated highway systems. In Proceedingsof the American Control Conference, pages pp. 653 – 657, Seattle,USA, 1995.

[CT00] Chieh Chen and Masayoshi Tomizuka. Lateral control of com-mercial heavy vehicles. Vehicle System Dynamics, (33):391–420,2000.

[DYT00] J. D. Demerly and K. Youcef-Toumi. Non-linear analysis of vehicledynamics (NAVDyn): A reduced order model for vehicle handlinganalysis. Technical report, SAE Paper 2000-01-1621, 2000.

[Edl91] S. Edlund. Tyre models: Subreport –91. Technical report, VolvoTruck Corporation, 1991. (classified).

[Ell69] John Ronaine Ellis. Vehicle Dynamics. Business Books Ltd.,London, 1969.

[Ell88] John R. Ellis. Road Vehicle Dynamics. John R. Ellis Inc, 1988.

[Ell94] J.R. Ellis. Vehicle Handling Dynamics. Mechanical EngineeringPublications Limited, London, 1994.

[FC93] Grant R. Fowles and George L. Cassiday. Analytical Mechanics.Saunders College Publishing, 5 edition, 1993.

[Gäf01] M. Gäfvert. Studies on yaw-control of heavy-duty trucks us-ing unilateral braking. Technical Report ISRN LUTFD2/TFRT–7598–SE, Department of Automatic Control, Lund Institute ofTechnology, Sweden, 2001. DICOSMOS Report.

[GSC00] M. Gäfvert, M. Sanfridsson, and V. Claesson. Truck modelfor yaw and roll dynamics control. Technical Report ISRNLUTFD2/TFRT–7588–SE, Department of Automatic Control,Lund Institute of Technology, 2000. DICOSMOS Report.

[KN00] U. Kiencke and L. Nielsen. Automotive Control Systems. Springer,2000.

[Leu70] Philip M. Leucht. The directional dynamics of the commercialtractor-semitrailer vehicle during braking. Technical Report SAEPaper 700371, Society of Automotive Engineers, 1970.

[LU84] P. Lidström and U. Uhlhorn. Teknisk mekanik 2f. Avdelningenför Mekanik, Lunds Tekniska Högskola, 1984. (In Swedish).

[Mik68] E. C. Mikulcik. The dynamics of Tractor-Semitrailer Vehicles: TheJackknifing Problem. PhD thesis, Cornell University, 1968.

[Mik71] E. C. Mikulcik. The dynamics of tractor-semitrailer vehicles: Thejackknifing problem. Technical Report SAE Paper No. 710045,Society of Automotive Engineers, 1971.

29

[RA95] R. Ranganathan and A. Aia. Development of heavy vehicledynamic stability analysis model using MATLAB/SIMULINK.Technical report, SAE Paper 952638, 1995.

[Say92] M. W. Sayers. Computer modeling and simulation of vehicledynamics. UMTRI Research Review, 23(1):1–15, 1992.

[SCG00] M. Sanfridson, V. Claesson, and M. Gäfvert. Investigation andrequirements of a computer control system in a heavy-duty truck.Technical Report TRITA-MMK 2000:5, ISSN 1400-1179, ISRNKTH/MMK–00/5–SE, Mechatronics Lab, Department of MachineDesign, Royal Institute of Technology, 2000. DICOSMOS Report.

[SPP96] D. J. Schuring, W. Pelz, and M. G. Pottinger. A model for combinedtire cornering and braking forces. Technical report, SAE Paper960180, 1996.

30

A. Nomenclature

Body configuration and reference systems

Btu Tractor unsprung body

Bsu Semitrailer unsprung body

Bts Tractor sprung body

Bss Semitrailer sprung body

Bh Hitch (tractor rear part) body

Stu Tractor reference system attached to B tu

Ssu Semitrailer reference system attached to Bsu

Sts Tractor reference system attached to B ts

Sss Semitrailer reference system attached to Bss

Sh Hitch wheel reference system attached to Bh

O Origin for Stu, Ssu, Sts, Sss, Sh

Indices

tu Tractor unsprung

su Semitrailer unsprung

ts Tractor sprung

ss Semitrailer sprung

h Hitch

f Front axle

r Rear axle

s Semitrailer axle

l Left side

r Right side

Geometry

rCM ,ts Center of mass location for B ts

rCM ,ss Center of mass location for Bss

rs,i, j Location of suspension at side j on axle i

rw,i, j Location of wheel at side j on axle i

31

Kinematics

vO velocity of O expressed in Stu

U Longitudinal component of vO

V Lateral component of vO

W Vertical component of vO

ω tu Angular velocity of Stu

r Yaw velocity of Stu

ω su Angular velocity of Ssu

r′ Yaw velocity of Ssu

ψ Articulation angle (angle between Stu and Ssu)χ t Tractor pitch angle

φ t Tractor roll angle

χs Semitrailer pitch angle

φs Semitrailer roll angle

φh Hitch body roll angle

Rtuts Coordinate transformation matrix from Sts to Stu

Rsuss Coordinate transformation matrix from Sss to Ssu

Rtuh Coordinate transformation matrix from Sh to Stu

Rsutu Coordinate transformation matrix from Ssu to Stu

Kinetics

mts Tractor mass (Bts)mss Semitrailer mass (Bss)Its Tractor inertial tensor with respect to O

Iss Semitrailer inertial tensor with respect to O

F′ts Internal force between B ts and Bh

M ′ts Internal moment between B ts and Bh

Fss Internal force between Bss and Bh

M ′ss Internal moment between Bss and Bh

Fts Sum of external forces acting on B ts

Mts Sum of external moments acting on B ts

Fss Sum of external forces acting on Bss

Mss Sum of external moments acting on Bss

Fs,i, j Suspension force at side j on axle i.

Fw,i, j Tire-road contact force at wheel on side j of axle i

32

B. Simulink Model

1

Out1

delta

lambda

state

mu

Wheels

t

delta

lambda

ScenarioClock

Fs

Ms

state

Chassis

mu

state

Fs

Ms

Axles

10{20} 10{20}

2{6}

3{9}

10

1515

1515

Figure 11 Top level view of Simulink model

2

Ms

1

Fs

mu_l

mu_r

state

Fs_l

Fs_r

Ms

Fw_l

Fw_r

phi

Semtrailer Axle 3

mu_l

mu_r

state

Fs_l

Fs_r

Ms

Fw_l

Fw_r

phi

Semtrailer Axle 2

mu_l

mu_r

state

Fs_l

Fs_r

Ms

Fw_l

Fw_r

phi

Semtrailer Axle 1

mu_l

mu_r

state

Fs_l

Fs_r

Ms

Fw_l

Fw_r

phi

Rear Axle

mu_l

mu_r

state

Fs_l

Fs_r

Ms

Fw_l

Fw_r

phi

Front Axle

2

state

1

mu

3

3

3

3

3

3

3

3

3

3

3

F_T

3

3Ms_f

3

3Ms_r

2

2<mu_fl>

2

2<mu_fr>

2

2

<mu_rl>

2

2

<mu_rr>

2

2

<mu_s1l>

2

2

<mu_s1r>

10{20}

2{6}

3{9}

15

15

15

15

15

15

3

3

3

3

3

3

3

33

3F_S

3

3

Ms_s

2

2

<mu_s2l>

2

2

<mu_s2r>

3

3

3

33

3

F_S

3

3

Ms_s

2

2

<mu_s3l>

2

2

<mu_s3r>

3

3

3

3

F_S

33

F_S

3

3

Ms_s

3

3

Ms_s

Figure 12 Axles block view of Simulink model

33

1

mu

delta

lambda

state

mu

alpha

front Left Wheel

delta

lambda

state

mu

alpha

Semitrailer 3 Right Wheel

delta

lambda

state

mu

alpha

Semitrailer 3 Left Whel2

delta

lambda

state

mu

alpha

Semitrailer 2 Right Wheel1

delta

lambda

state

mu

alpha

Semitrailer 2 Left Wheel

delta

lambda

state

mu

alpha

Semitrailer 1 Right Wheel

delta

lambda

state

mu

alpha

Semitrailer 1 Left Wheel

delta

lambda

state

mu

alpha

Rear Right Wheel

delta

lambda

state

mu

alpha

Rear Left Wheel

delta

lambda

state

mu

alpha

Front Right Wheel

Demux

3

state

2

lambda

1

delta

2

2

mu_fl

2

2

mu_fr

2

2

mu_rl

2

2

mu_rr

2

2

mu_s1l

2

2mu_s1r

10{20}

10

15

15

15

15

15

15

15

15

15

1515

2mu_s2l

22

mu_s2r

2 2

mu_s3l

22

mu_s3r

Figure 13 Wheels block view of Simulink model

34

C. Moments of inertia

It is convenient to consider the body mass distribution as combinationsof rectangular blocks with homogeneous density. This makes it easy tocompute the moments of inertia for the complete chassis body with respectto a certain origin and a certain coordinate system aligned with the block.Since the moments of inertia are additive, the resulting moments of inertiafor the body are the sum of the moments of inertia for the blocks. Recallthat the inertia terms occurring in the model equations are Ii j =

∫rir j dmP

with i, j ∈ {x, y, z}, instead of the traditional terms defined as the elementsof the matrix

∫ (rT

P rP1− rP rTP

)dmP.

Now regard a homogeneous rectangular block of mass m with the sidesA, B, C in the ex, ey, ez directions respectively. The position vector of the cgof the block with respect to O is rck = Rx ex + Ry ey + Rz ez. Then

Ixx = m(

R2x + A2

12

)Ix y = mRx Ry Ixz = mRx Rz

Iyx = Ix y Iy y = m(

R2y + B2

12

)Iyz = mRy Rz

Izx = Ixz Izy = Iyz Izz = m(

R2z + C2

12

)(59)

35

D. Validation Results

D.1 Lane-change maneuver (i)

0 1 2 3 4 5 6 7−1

−0.5

0

0.5

1

1.5Wheel angle δ

[deg

]

0 1 2 3 4 5 6 750

60

70

80

90

100

110Vehicle speed U

[km

/h]

0 1 2 3 4 5 6 7−2

−1.5

−1

−0.5

0

0.5

1

1.5

2Kingpin angle ψ

[deg

]

t [s]0 1 2 3 4 5 6 7

−6

−4

−2

0

2

4

6Yaw rate r

[deg

/s]

t [s]

0 1 2 3 4 5 6 7−10

−5

0

5

10

15Suspension travel: front left

[mm

]

0 1 2 3 4 5 6 7−10

−5

0

5

10

15Suspension travel: front right

[mm

]

0 1 2 3 4 5 6 7−10

−5

0

5

10

15Suspension travel: rear left

[mm

]

t [s]0 1 2 3 4 5 6 7

−10

−5

0

5

10Suspension travel: rear right

[mm

]

t [s]

Figure 14 Solid: Experiment; Dashed: 9-DOF model; Dotted: Linear 4-DOFmodel

36

D.2 Lane-change maneuver (ii)

0 1 2 3 4 5 6−3

−2

−1

0

1

2

3Wheel angle δ

[deg

]0 1 2 3 4 5 6

50

60

70

80

90

100

110Vehicle speed U

[km

/h]

0 1 2 3 4 5 6−5

−4

−3

−2

−1

0

1

2

3

4Kingpin angle ψ

[deg

]

t [s]0 1 2 3 4 5 6

−15

−10

−5

0

5

10

15Yaw rate r

[deg

/s]

t [s]

0 1 2 3 4 5 6−15

−10

−5

0

5

10

15


[mm

]

0 1 2 3 4 5 6−15

−10

−5

0

5

10

15


[mm

]

0 1 2 3 4 5 6−15

−10

−5

0

5

10


[mm

]

t [s]0 1 2 3 4 5 6

−15

−10

−5

0

5

10


[mm

]

t [s]


37

D.3 Step-steering maneuver (i)

0 2 4 6 8−1.4

−1.2

−1

−0.8

−0.6

−0.4

−0.2

0

0.2Wheel angle δ

[deg

]

0 2 4 6 850

60

70

80

90

100

110Vehicle speed U

[km

/h]

0 2 4 6 8−2.5

−2

−1.5

−1

−0.5

0

0.5Kingpin angle ψ

[deg

]

t [s]0 2 4 6 8

−5

−4

−3

−2

−1

0

1Yaw rate r

[deg

/s]

t [s]

0 2 4 6 8−10

−5

0


[mm

]

0 2 4 6 8−2

0

2

4

6

8

10

12


[mm

]

0 2 4 6 8−10

−8

−6

−4

−2

0

2


[mm

]

t [s]0 2 4 6 8

−2

0

2

4

6

8

10


[mm

]

t [s]


38

D.4 Step-steering maneuver (ii)

0 1 2 3 4−3

−2.5

−2

−1.5

−1

−0.5

0

0.5Wheel angle δ

[deg

]0 1 2 3 4

50

60

70

80

90

100

110Vehicle speed U

[km

/h]

0 1 2 3 4−6

−5

−4

−3

−2

−1

0

1Kingpin angle ψ

[deg

]

t [s]0 1 2 3 4

−14

−12

−10

−8

−6

−4

−2

0

2Yaw rate r

[deg

/s]

t [s]

0 1 2 3 4−30

−25

−20

−15

−10

−5

0


[mm

]

0 1 2 3 40

5

10

15

20

25

30


[mm

]

0 1 2 3 4−15

−10

−5

0

5


[mm

]

t [s]0 1 2 3 4

0

5

10

15

20

25

30


[mm

]

t [s]


39

D.5 Random steering maneuvers

0 10 20 30 40−3

−2

−1

0

1

2

3Wheel angle δ

[deg

]

0 10 20 30 4050

60

70

80

90

100

110Vehicle speed U

[km

/h]

0 10 20 30 40−5

0

5Kingpin angle ψ

[deg

]

t [s]0 10 20 30 40

−15

−10

−5

0

5

10

15Yaw rate r

[deg

/s]

t [s]

0 10 20 30 40−20

−15

−10

−5

0

5

10

15


[mm

]

0 10 20 30 40−20

−15

−10

−5

0

5

10

15


[mm

]

0 10 20 30 40−25

−20

−15

−10

−5

0

5

10


[mm

]

t [s]0 10 20 30 40

−20

−15

−10

−5

0

5

10

15


[mm

]

t [s]


40

E. Simulation Results

E.1 Lane-Change Maneuver

0 1 2 3 4 5 6 7 8 921.6

21.8

22

22.2

22.4

t [s]

U [m

/s]

Longitudinal translation

0 1 2 3 4 5 6 7 8 9−1.5

−1

−0.5

0

0.5

1

1.5

t [s]

V [m

/s]

Lateral translation

0 1 2 3 4 5 6 7 8 9−1.5

−1

−0.5

0

0.5

t [s]

W [m

/s],

Z [m

]

Vertical translation

Figure 19 Longitudinal, lateral, and heave motion. (Solid: position; Dashed: ve-locity)

41

0 1 2 3 4 5 6 7 8 9−20

−15

−10

−5

0

5

10

15

20

t [s]

r [d

eg/s

]

Yaw

0 1 2 3 4 5 6 7 8 9−20

−15

−10

−5

0

5

10

15

t [s]

ψ . [d

eg/s

], ψ

[deg

]

Articulation

Figure 20 Yaw motion. (Solid: position; Dashed: velocity)

0 1 2 3 4 5 6 7 8 9−5

0

5

t [s]

φ. 1 [deg

/s],

φ 1 [deg

]

Tractor roll

0 1 2 3 4 5 6 7 8 9−10

−5

0

5

t [s]

χ. 1 [deg

/s],

χ 1 [deg

]

Tractor pitch

0 1 2 3 4 5 6 7 8 9−5

0

5

t [s]

φ. 2 [deg

/s],

φ 2 [deg

]

Semitrailer roll

0 1 2 3 4 5 6 7 8 9−2

0

2

4

t [s]

χ. 2 [deg

/s],

χ 2 [deg

]

Semitrailer pitch

Figure 21 Roll and pitch motion. (Solid: position; Dashed: velocity)

42

0 1 2 3 4 5 6 7 8 9−4

−3

−2

−1

0

1

2

3

4

t [s]

δ [d

eg]

Front steering angle

0 1 2 3 4 5 6 7 8 9−0.6

−0.4

−0.2

0

0.2

0.4

0.6

t [s]

[g]

Kingpin lateral acceleration

Figure 22 Steering input and lateral acceleration at kingpin.

0 2 4 6 8 10−1

0

1

t [s]

λ fl

0 2 4 6 8 10−1

0

1

t [s]

λ fr

0 2 4 6 8 10−1

0

1

t [s]

λ rl

0 2 4 6 8 10−1

0

1

t [s]

λ rr

0 2 4 6 8 10−1

0

1

t [s]

λ s1l

0 2 4 6 8 10−1

0

1

t [s]

λ s1r

0 2 4 6 8 10−1

0

1

t [s]

λ s2l

0 2 4 6 8 10−1

0

1

t [s]

λ s2r

0 2 4 6 8 10−1

0

1

t [s]

λ s3l

0 2 4 6 8 10−1

0

1

t [s]

λ s3r

Figure 23 Longitudinal slip inputs. (Zero since no braking is applied.)

43

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,fl [N

]

0 2 4 6 8 10−5

0

5x 10

4

t [s]

Fw

,fr [N

]

0 2 4 6 8 10−1

0

1x 10

5

t [s]

Fw

,rl [N

]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,rr [N

]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,s1l

[N]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,s1r

[N]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,s2l

[N]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,s2r

[N]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,s3l

[N]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,s3r

[N]

Figure 24 Tire forces in tractor and semitrailer unsprung reference systems re-spectively. (Dash-dotted: longitudinal; Dashed: lateral; Solid: vertical)

44


45

E.2 Lane-Change Maneuver with Jack-knifing

0 1 2 3 4 5 6 7 8 9−20

−10

0

10

20

30

t [s]

U [m

/s]

Longitudinal translation

0 1 2 3 4 5 6 7 8 9−25

−20

−15

−10

−5

0

5

t [s]

V [m

/s]

Lateral translation

0 1 2 3 4 5 6 7 8 9−1.5

−1

−0.5

0

0.5

t [s]

W [m

/s],

Z [m

]

Vertical translation

Figure 26 Longitudinal, lateral, and heave motion. (Solid: position; Dashed: ve-locity)

46

0 1 2 3 4 5 6 7 8 90

10

20

30

40

50

60

70

80

90

t [s]

r [d

eg/s

]

Yaw

0 1 2 3 4 5 6 7 8 9−50

0

50

100

150

200

t [s]

ψ . [d

eg/s

], ψ

[deg

]

Articulation

Figure 27 Yaw motion. (Solid: position; Dashed: velocity)

0 1 2 3 4 5 6 7 8 9−20

−10

0

10

20

t [s]

φ. 1 [deg

/s],

φ 1 [deg

]

Tractor roll

0 1 2 3 4 5 6 7 8 9−10

0

10

20

t [s]

χ. 1 [deg

/s],

χ 1 [deg

]

Tractor pitch

0 1 2 3 4 5 6 7 8 9−20

−10

0

10

20

t [s]

φ. 2 [deg

/s],

φ 2 [deg

]

Semitrailer roll

0 1 2 3 4 5 6 7 8 9−2

0

2

4

t [s]

χ. 2 [deg

/s],

χ 2 [deg

]

Semitrailer pitch

Figure 28 Roll and pitch motion. (Solid: position; Dashed: velocity)

47

0 1 2 3 4 5 6 7 8 9−4

−3

−2

−1

0

1

2

3

4

t [s]

δ [d

eg]

Front steering angle

0 1 2 3 4 5 6 7 8 9−1

−0.5

0

0.5

1

t [s]

[g]

Kingpin lateral acceleration

Figure 29 Steering input and lateral acceleration at kingpin.

0 2 4 6 8 10−1

0

1

t [s]

λ fl

0 2 4 6 8 10−1

0

1

t [s]

λ fr

0 2 4 6 8 100

0.5

1

t [s]

λ rl

0 2 4 6 8 100

0.5

1

t [s]

λ rr

0 2 4 6 8 10−1

0

1

t [s]

λ s1l

0 2 4 6 8 10−1

0

1

t [s]

λ s1r

0 2 4 6 8 10−1

0

1

t [s]

λ s2l

0 2 4 6 8 10−1

0

1

t [s]

λ s2r

0 2 4 6 8 10−1

0

1

t [s]

λ s3l

0 2 4 6 8 10−1

0

1

t [s]

λ s3r

Figure 30 Longitudinal slip inputs.

48

0 2 4 6 8 10−1

0

1x 10

5

t [s]

Fw

,fl [N

]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,fr [N

]

0 2 4 6 8 10−2

0

2x 10

5

t [s]

Fw

,rl [N

]

0 2 4 6 8 10−1

0

1

2x 10

5

t [s]

Fw

,rr [N

]0 2 4 6 8 10

−1

0

1x 10

5

t [s]

Fw

,s1l

[N]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,s1r

[N]

0 2 4 6 8 10−1

0

1x 10

5

t [s]

Fw

,s2l

[N]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,s2r

[N]

0 2 4 6 8 10−1

0

1x 10

5

t [s]

Fw

,s3l

[N]

0 2 4 6 8 10−5

0

5

10x 10

4

t [s]

Fw

,s3r

[N]

Figure 31 Tire forces in tractor and semitrailer unsprung reference systems re-spectively. (Dash-dotted: longitudinal; Dashed: lateral; Solid: vertical)

49


50

a 9-dof tractor-semitrailer dynamic handling model for...

Documents